Integrating prior knowledge with a software reliability growth model

The authors describe an application of Ohba´s inflection S reliability growth model to an IBM software product consisting of 1200 KLOC. The central theme deals with the management requirement that forecasts be made early in the test cycle when there is little data. A simple brute force Monte Carlo simulation is implemented which utilizes prior knowledge about the total number of failures in the product prior to test. Attention is also given to the decisions made regarding the data to select as input to the model. Scaled forecasts and field results are presented which show reasonably good results